Deploy tested AI components
Connect
SPIRITT
GriptapeOperate your AI applications
Take runs through to results
Build and operate Griptape assistants, knowledge bases and deployments using your private sources and execution history. Connect approved document sources to knowledge bases and verify ingestion before using the results.
What you can do with Griptape.
Private knowledge in service
Connect approved document sources to knowledge bases and verify ingestion before using the results.
Runs with an owner
Follow assistant runs through status, logs and error details rather than stopping at a queued request.
Deployed building blocks
Ship repository-backed tools and structures with deployment checks and clear operational records.
Works with your other apps.
Get started in three steps.
01
Open your workspace
Start a SPIRITT workspace for private knowledge in service and give it your operating priorities.
02
Connect Griptape
Connect your Griptape account and choose the records and actions this workspace may use. Define source access, run budgets and deployment scope.
03
Delegate an outcome
Ask SPIRITT to launch a knowledge assistant. It carries the work through and checks the result.
Hand over a mission.
Choose a mission. Make it yours.
Put Griptape to work.
Take runs through to results
Questions
Griptape, with SPIRITT.
01Can SPIRITT build Griptape assistants?+
Yes. It can create assistants and attach supported models, knowledge bases, rulesets, structures and tools, then run and inspect conversations.
02Does starting a run mean it finished?+
No. SPIRITT checks run status and available execution records before reporting an output or failure.
03Can it connect private documents?+
It can configure data connectors, knowledge bases and ingestion jobs within the authorized source boundary, then test retrieval.
04Can it deploy code from GitHub?+
Repository-backed deployments are available for tools, functions and structures. The relevant deployment status or history supplies the release evidence.
05Can it control usage?+
It can inspect usage limits, current credits and execution metrics, then operate your alert thresholds. Purchasing additional capacity is a separate decision.
06Can it retry failures safely?+
It can inspect errors and request an eligible retry. If retry is unavailable, the returned current state is not represented as a new successful run.